技术领域technical field
本发明涉及道路边界检测技术，更具体地，涉及一种基于激光点云的道路边界检测方法。The present invention relates to road boundary detection technology, more specifically, to a road boundary detection method based on laser point cloud.
背景技术Background technique
近年来，智能驾驶技术受到国际IT企业和汽车行业的广泛关注，也成为未来最值得期待的汽车技术之一。道路信息检测是智能驾驶技术中的关键技术，道路边界检测是道路信息检测中的重点和难点。在交通环境中，道路边界用来划分道路可行驶区域和非行驶区域，同时凸起的道路边界也是车辆行驶过程中的障碍物。车辆检测出道路边界不仅可以保证车辆在左右道路边界内行驶，而且可以辅助进行道路趋势的预测，辅助车辆在行驶过程中进行路径规划。In recent years, intelligent driving technology has attracted widespread attention from international IT companies and the automotive industry, and has become one of the most anticipated automotive technologies in the future. Road information detection is a key technology in intelligent driving technology, and road boundary detection is the key and difficult point in road information detection. In the traffic environment, the road boundary is used to divide the drivable area and the non-drivable area of the road, and the raised road boundary is also an obstacle during the driving process of the vehicle. The detection of the road boundary by the vehicle can not only ensure that the vehicle travels within the left and right road boundaries, but also assist in the prediction of the road trend and assist the vehicle in path planning during driving.
现有的技术方案有：The existing technical solutions are:
方案1：Chen T,Dai B,Liu D.Velodyne-based curb detection up to 50metersaway[C]//Intelligent Vehicles Symposium(IV),USA：IEEE， 2015:241-248。该论文提出了一种远距离检测道路边界的方法，该方法首先获取每一条激光扫描线，然后根据距离判别法和霍夫变换法提取道路边界点，最后将提取道的特征点作为种子点，通过高斯迭代回归过程拟合出道路边界。Scheme 1: Chen T, Dai B, Liu D. Velodyne-based curb detection up to 50metersaway[C]//Intelligent Vehicles Symposium(IV), USA: IEEE, 2015:241-248. This paper proposes a method for long-distance detection of road boundaries. This method first obtains each laser scanning line, then extracts road boundary points according to the distance discrimination method and Hough transform method, and finally uses the feature points of the extracted road as seed points. The road boundaries are fitted by a Gaussian iterative regression process.
方案2：公开号为CN10485083，发明名称为“基于三维激光雷达的道路边界检测方法”的中国专利申请提供了三维数据栅格化检测道路边界的方法。该方法首先将点云数据进行栅格化处理，然后将栅格图转化为距离灰度图，通过阈值填充法获取道路的整体轮廓，然后通过区域生长法和栅格图相结合获取可行区域轮廓图，最后对可行区域轮廓图进行二次函数拟合获取道路边界。Solution 2: The Chinese patent application with the publication number CN10485083 and the title of the invention entitled "Road Boundary Detection Method Based on 3D LiDAR" provides a method for detecting road boundaries with 3D data rasterization. This method first rasterizes the point cloud data, then converts the raster image into a distance grayscale image, obtains the overall outline of the road through the threshold filling method, and then obtains the feasible area outline by combining the region growing method and the raster image Finally, quadratic function fitting is performed on the contour map of the feasible area to obtain the road boundary.
上述两个技术方案中，对道路表面存在障碍物和道路边界上存在遮挡这一实际问题涉及较少。方案1中通过提取道路激光扫描线来对道路边界点进行检测，当道路平面上存在障碍物遮挡时，激光扫描线无法扫描到道路边界，这种情况就会导致道路边界点丢失，最终无法从激光束扫描线中获取道路边界点。方案2虽然考虑了道路表面存在障碍物的情况，但是没有考虑道路边界上存在绿化带等遮挡导致原始点云中道路边界的结构发生无规则变化这一实际情况，而这一情况会导致方案2中对道路边界点的检测存在较大偏差。Among the above two technical solutions, the actual problems of obstacles on the road surface and occlusions on the road boundary are less involved. In scheme 1, the road boundary points are detected by extracting the road laser scanning lines. When there are obstacles on the road plane, the laser scanning lines cannot scan the road boundaries. The road boundary points are obtained from the laser beam scanning line. Although Scheme 2 considers the existence of obstacles on the road surface, it does not consider the fact that there are green belts and other occlusions on the road boundary that lead to irregular changes in the structure of the road boundary in the original point cloud, and this situation will lead to Scheme 2 There is a large deviation in the detection of road boundary points.
发明内容Contents of the invention
针对道路表面存在障碍物和道路边界上存在绿化带遮挡等难点，本发明一种基于激光点云的道路边界检测方法，包括：Aiming at difficulties such as obstacles on the road surface and green belt occlusions on the road boundary, the present invention provides a road boundary detection method based on laser point cloud, including:
S10，对道路点云数据进行预处理，去除道路点云中的噪声；S10, preprocessing the road point cloud data to remove noise in the road point cloud;
S20，根据车辆的行驶轨迹获取道路初始横切面法向量，获取完整的道路横切面；S20, obtaining the normal vector of the initial cross-section of the road according to the driving trajectory of the vehicle, and obtaining a complete cross-section of the road;
S30，通过单线点云获取道路初始横切面的轮廓；S30, obtaining the outline of the initial cross-section of the road through the single-line point cloud;
S40，根据道路边界的特征，通过窗口法检测道路边界点；S40, according to the characteristic of road boundary, detect road boundary point by window method;
S50，拟合道路边界点获取完整的道路边界线。S50, fitting the road boundary points to obtain a complete road boundary line.
本发明能够快速准确检测实际道路表面存在障碍物和道路边界上存在绿化带遮挡时的道路边界。本发明可用于智能驾驶技术和自动驾驶级别的高精度地图制作技术。本发明能够辅助车辆确定道路的可行驶域和非行使域，保证车辆的安全驾驶，同时本发明也能用于3D 高精度地图制图中划定道路区域的工作。The invention can quickly and accurately detect the road boundary when obstacles exist on the actual road surface and the road boundary is blocked by green belts on the road boundary. The invention can be used in intelligent driving technology and automatic driving level high-precision map making technology. The invention can assist the vehicle to determine the drivable area and the non-drivable area of the road to ensure the safe driving of the vehicle, and at the same time, the invention can also be used for demarcating road areas in 3D high-precision map drawing.
本发明与现有的技术相比，其显著的优点为：Compared with the prior art, the present invention has the remarkable advantages of:
(1)本发明中通过将道路初始横切面前后一定范围点云向其本身投影的方法提取道路的横切面，这样可以提高待检测道路边界的完整性,促使检测结果更加准确。(1) In the present invention, the cross-section of the road is extracted by projecting a certain range of point clouds before and after the initial cross-section of the road to itself, which can improve the integrity of the road boundary to be detected and promote more accurate detection results.
(2)由于道路上表面的形态特征容易因障碍物的出现而发生变化，而道路下表面的形态特征不会改变，本发明通过设计单线点云从道路平面下方获取道路横切面轮廓，这种方法对于道路中存在障碍物和道路边界上存在遮挡的情况时仍然会有良好的道路边界检测结果。(2) Since the morphological features of the upper surface of the road are easy to change due to the appearance of obstacles, but the morphological features of the lower surface of the road will not change, the present invention obtains the road cross-section profile from below the road plane by designing a single-line point cloud. The method still has good road boundary detection results when there are obstacles in the road and occlusions on the road boundary.
(3)本发明时间复杂度低，鲁棒性好，能实时获取道路边界检测的结果。(3) The present invention has low time complexity and good robustness, and can obtain the result of road boundary detection in real time.
附图说明Description of drawings
图1为本发明的方法的一个实施方式的流程图；Fig. 1 is the flowchart of an embodiment of the method of the present invention;
图2为图1中的道路横切面获取步骤的流程图；Fig. 2 is the flow chart of the road cross-section acquisition step in Fig. 1;
图3为道路模型；Fig. 3 is a road model;
图4为第1级搜索点搜索方法的示意图；Fig. 4 is a schematic diagram of the first-level search point search method;
图5为搜索点球形邻域的示意图；Fig. 5 is the schematic diagram of searching point spherical neighborhood;
图6二级搜索点搜索策略的示意图；Fig. 6 is a schematic diagram of a secondary search point search strategy;
图7为图1中的道路横切面提取步骤的流程图；Fig. 7 is the flow chart of the road cross-section extraction step in Fig. 1;
图8为道路轮廓提取的示意图；Fig. 8 is a schematic diagram of road contour extraction;
图9A-9B为窗口法获取道路边界位置的示意图；9A-9B are schematic diagrams of obtaining the position of the road boundary by the window method;
图10为无植被遮挡的道路边界检测的结果示意图；Fig. 10 is a schematic diagram of the result of road boundary detection without vegetation occlusion;
图11为植被遮挡的道路边界检测的结果示意图；Fig. 11 is a schematic diagram of the result of road boundary detection covered by vegetation;
图12为被汽车遮挡的道路边界检测的结果示意图；Fig. 12 is a schematic diagram of the results of road boundary detection blocked by cars;
图13A-13B为无植被遮挡的道路边界检测的结果示意图；13A-13B are schematic diagrams of the results of road boundary detection without vegetation occlusion;
图14A-14B为植被遮挡的道路边界检测的结果示意图；14A-14B are schematic diagrams of the results of road boundary detection covered by vegetation;
图15A-15B为汽车遮挡的道路边界检测的结果示意图。15A-15B are schematic diagrams of the results of road boundary detection for car occlusion.
具体实施方式detailed description
下面参照附图描述本发明的实施方式，其中相同的部件用相同的附图标记表示。Embodiments of the present invention are described below with reference to the drawings, in which like parts are denoted by like reference numerals.
图1为本发明的方法的一个实施方式的流程图。Figure 1 is a flow chart of one embodiment of the method of the present invention.
在S10中，对道路点云数据进行预处理，去除道路点云中的噪声。In S10, the road point cloud data is preprocessed to remove noise in the road point cloud.
获取的道路点云中会存在明显离群点，离群点往往由测量噪声引入，其特征是在空间中分布稀疏，信息量较小，属于无用信息，所以离群点表达的信息可以忽略不计，所以可以通过滤波方法对离群点进行滤除，提高整体数据的质量，常用的滤波方法有统计滤波，半径滤波方法，但不限于这些方法，对原始点云滤波处理后获取的数据为滤波后的数据。There will be obvious outliers in the obtained road point cloud. Outliers are often introduced by measurement noise. They are characterized by sparse distribution in space and small amount of information. They are useless information, so the information expressed by outliers can be ignored. , so the outlier points can be filtered out by the filtering method to improve the quality of the overall data. Commonly used filtering methods include statistical filtering and radius filtering methods, but are not limited to these methods. The data obtained after filtering the original point cloud is filtering after the data.
常见道路原始点云量非常庞大，为了对道路点云进行快速有序的处理，本发明对滤波后的数据进行条带型切割，之后对条带型数据按照条带顺序进行道路边界提取，这样可以使搜索算法的整个搜索域由全部点云转化为单个的分割后的条带数据，大大的提高搜索算法的运行效率。通过分割处理可以获得分割后的条带型数据。The amount of raw point clouds of common roads is very large. In order to process the road point clouds quickly and orderly, the present invention cuts the filtered data into strips, and then extracts the road boundaries from the striped data according to the order of the strips. The entire search domain of the search algorithm can be converted from all point clouds to a single segmented strip data, which greatly improves the operating efficiency of the search algorithm. Split strip data can be obtained through split processing.
在S20，根据车辆的行驶轨迹获取道路初始横切面法向量，之后采用“由粗到精”的搜索策略获取完整的道路横切面。In S20, the normal vector of the initial cross-section of the road is obtained according to the driving trajectory of the vehicle, and then the complete road cross-section is obtained by using the "coarse to fine" search strategy.
在S30，通过单线点云获取道路横切面的轮廓。At S30, the outline of the cross-section of the road is acquired through the single-line point cloud.
在S40，根据道路边界的特征，通过窗口法检测道路边界点。At S40, the road boundary points are detected by the window method according to the characteristics of the road boundary.
在S50，拟合道路边界点获取完整的道路边界线。At S50, the road boundary points are fitted to obtain a complete road boundary line.
为方便理解本发明，图2显示了道路的模型。如图2所示，该道路的模型分为道路区域和非道路区域，图中灰色平面代表道路初始横切面，该平面前后一定范围内点云没有向其进行投影，所以该平面上不存在数据；若该平面前后一定范围内点云向道路初始横切面进行投影，则投影后的平面称作道路横切面。道路初始横切面中的虚线代表道路横切面轮廓，反映道路的结构。道路平面和道路凸起处的交线为道路边界线。道路边界线上的点为道路边界点。在每一个道路初始横切面的上边缘存在一个车辆轨迹点，代表车辆行驶过的位置。相邻两个车辆轨迹点之间的箭头代表车轮的运动方向，为了简要言明图意，图2中只显示首尾道路初始横切面。To facilitate the understanding of the present invention, Fig. 2 shows a road model. As shown in Figure 2, the road model is divided into road area and non-road area. The gray plane in the figure represents the initial cross section of the road. The point cloud within a certain range before and after this plane is not projected to it, so there is no data on this plane. ; If the point cloud within a certain range before and after the plane is projected to the initial cross-section of the road, the projected plane is called the road cross-section. The dotted line in the initial cross section of the road represents the profile of the road cross section, reflecting the structure of the road. The intersection line between the road plane and the road bump is the road boundary line. Points on the road boundary line are road boundary points. There is a vehicle trajectory point on the upper edge of each initial cross-section of the road, which represents the location where the vehicle has traveled. The arrows between two adjacent vehicle trajectory points represent the direction of movement of the wheels. In order to briefly explain the intention, only the initial cross-section of the front and rear roads is shown in Figure 2.
如图3所示，步骤S20可以进一步细化为：As shown in Figure 3, step S20 can be further refined as:
S201，获取道路横切面法向量。S201. Obtain a road cross-section normal vector.
根据相邻的2个车辆轨迹点的坐标获取2个车辆轨迹点之间的道路横切面的法向量：Obtain the normal vector of the road cross-section between two vehicle trajectory points according to the coordinates of two adjacent vehicle trajectory points:
(xi,yi,zi)和(xi+1,yi+1,zi+1)分别是第i个和第i+1个轨迹点的坐标， (xid,yid,zid)是2个车辆轨迹点之间初始横切面的法向量，本发明中道路横切面的法向量和相邻轨迹点的运动方向相关，该步骤获取了道路初始横切面向量(xid,yid,zid)。(xi ,yi ,zi ) and (xi+1 ,yi+1 ,zi+1 ) are the coordinates of the i-th and i+1-th trajectory points respectively, (xid ,yid , zid ) is the normal vector of the initial cross-section between two vehicle trajectory points. In the present invention, the normal vector of the road cross-section is related to the direction of motion of the adjacent trajectory points. This step obtains the initial road cross-section vector (xid , yid , zid ).
S202，获取初始道路横切面范围。S202. Obtain an initial road cross-section range.
道路初始横切面即是未经道路初始横切面前后一定范围内点云向其投影的道路横切面。The initial cross-section of the road is the road cross-section projected to it by the point cloud within a certain range before and after the initial cross-section of the road.
1)设定道路初始横切面上边缘和下边缘。1) Set the upper and lower edges of the initial cross-section of the road.
本实施例中上边缘的高度计算公式为：The formula for calculating the height of the upper edge in this embodiment is:
Hup＝zi-Hl+Hi (2)Hup =zi -Hl +Hi (2)
本实施例中下边缘的高度计算公式为：The formula for calculating the height of the lower edge in this embodiment is:
Hdown＝zi-Hl-Hi (3)Hdown ＝zi -Hl -Hi (3)
其中zi表示激光器在第i个轨迹点坐标的z值，Hl为激光器距离地面的高度，Hi为道路初始横切面上边缘距离地面的高度。Where zi represents the z value of the laser at the coordinates of the ith track point, Hl is the height of the laser from the ground, and Hi is the height of the edge of the initial cross-section of the road from the ground.
道路初始横切面的大小设定在一定范围内，过大会降低算法的执行效率，过小则不能跨越整条道路的宽度。在本发明中，通过设定道路初始横切面的范围约束点来确定道路初始横切面的位置和大小，如图4所示。The size of the initial cross-section of the road is set within a certain range. If it is too large, the execution efficiency of the algorithm will be reduced, and if it is too small, it will not be able to span the width of the entire road. In the present invention, the position and size of the initial road cross-section are determined by setting the range constraint points of the initial road cross-section, as shown in FIG. 4 .
2)设定道路初始横切面范围约束点。2) Set the constraint points of the initial cross-section range of the road.
本发明实施例中设定了4个范围点来确定道路初始横切面的范围，范围点的设定方式存在多种方式，道路初始横切面的范围设定和所有范围点的共同组成的区域有关。In the embodiment of the present invention, four range points are set to determine the range of the initial cross-section of the road. There are many ways to set the range points. The range setting of the initial cross-section of the road is related to the area composed of all the range points. .
首先获取道路横切面在非z方向的斜率kid：First obtain the slope kid of the road cross-section in the non-z direction:
道路初始切面左边界范围点的x坐标：The x-coordinate of the left boundary range point of the initial section of the road:
道路初始横切面右边界范围点的x坐标：The x-coordinate of the right boundary range point of the initial cross-section of the road:
道路初始横切面左边界范围点的y坐标：The y coordinate of the left boundary range point of the initial cross section of the road:
yleft＝-kid*(xleft-xi)+yi (7)yleft ＝-kid *(xleft -xi )+yi (7)
道路初始横切面右边界范围点的y坐标：The y-coordinate of the right boundary range point of the initial cross-section of the road:
yright＝-kid*(xright-xi)+yi (8)yright ＝-kid *(xright -xi )+yi (8)
公式中Eleft和Eright分别为从道路轨迹点开始，道路初始横切面沿着(xid,yid)方向的分别向左方向和右方向延伸的距离，最终4个范围约束点为(xright,yright,Hup)，(xright,yright,Hdown)，(xleft,yleft,Hup)和(xleft,yleft,Hdown)， 4个范围约束点如图4所示。In the formula, Eleft and Eright are respectively starting from the road trajectory point, and the initial cross-section of the road along the (xid , yid ) direction extends to the left and right respectively, and the final four range constraint points are (xright , yright , Hup ), (xright , yright , Hdown ), (xleft , yleft , Hup ) and (xleft , yleft , Hdown ), the four range constraint points are shown in Figure 4 shown.
S203，获取初始横切面上空非空邻域搜索点集。S203. Obtain an empty and non-empty neighborhood search point set on the initial cross section.
确定道路初始横切面的范围约束点后，需要确定道路初始横切面附近范围内道路数据的分布情况。在道路初始横切面前后范围内，点云主要集中在道路平面和道路中物体上，所以道路初始横切面上有些区域的前后范围内不存在点云。After determining the range constraint points of the initial cross-section of the road, it is necessary to determine the distribution of road data in the vicinity of the initial cross-section of the road. In the range before and after the initial cross-section of the road, the point cloud is mainly concentrated on the road plane and the objects in the road, so there are no point clouds in the front and rear range of some areas on the initial cross-section of the road.
优选地，采用“由粗到精”的搜索策略，这是获取道路横切面的关键，“由粗到精”指在道路初始横切面上选取多级搜索点，从第1级到第n级搜索点的邻域半径逐渐缩小(n>1)，但是搜索点的数量逐渐增加，所有搜索点的邻域能够覆盖道路及道路中所有物体分布的位置。通过“由粗到精”的搜索策略可以逐步去除道路初始横切面前后范围内不存在三维数据的区域，保留道路初始横切面前后存在原始点云的区域，本发明实施例中n值为3。Preferably, the "coarse to fine" search strategy is adopted, which is the key to obtaining the cross-section of the road. "From coarse to fine" refers to selecting multi-level search points on the initial cross-section of the road, from the first level to the nth level The neighborhood radius of the search point is gradually reduced (n>1), but the number of search points is gradually increased, and the neighborhood of all search points can cover the road and the location of all objects in the road. Through the "coarse to fine" search strategy, the areas without 3D data in the front and back of the initial cross section of the road can be gradually removed, and the areas with original point clouds in the front and back of the initial cross section of the road can be retained. In the embodiment of the present invention, the value of n is 3.
1)如图4所示，灰色平面为道路初始横切面，道路初始横切面上每个点代表一个搜索域的中心，搜索域分布在道路初始横切面的两侧，如图5所示，本实施例中使用的是球形搜索域，道路初始横切面将球形邻域分割成了相等的两部分，球形邻域中的点代表球形邻域搜索到的点云数据。同时，图4中道路初始横切面上均匀分布的每个点也是第1级搜索点，每一个第1级搜索点代表一个球形邻域的球心，每个球代表以第1级搜索点为中心的球形邻域。如果球形邻域中含有点云，则称该球邻域为非空邻域，称非空邻域中的球心为非空邻域搜索点，第1级搜索点的计算公式如下：1) As shown in Figure 4, the gray plane is the initial cross-section of the road. Each point on the initial cross-section of the road represents the center of a search domain, and the search domain is distributed on both sides of the initial cross-section of the road. As shown in Figure 5, this In the embodiment, a spherical search domain is used, and the initial road cross section divides the spherical neighborhood into two equal parts, and the points in the spherical neighborhood represent the point cloud data searched in the spherical neighborhood. At the same time, each point evenly distributed on the initial cross-section of the road in Figure 4 is also a first-level search point. Each first-level search point represents the center of a spherical neighborhood, and each sphere represents the first-level search point. The spherical neighborhood of the center. If the spherical neighborhood contains a point cloud, the spherical neighborhood is called a non-empty neighborhood, and the center of the sphere in the non-empty neighborhood is called a non-empty neighborhood search point. The calculation formula of the first-level search point is as follows:
其中N1和N2为横切面在横向和纵向的搜索点的间隔数量。 (x(i,j),y(i,j),z(i,j))是位于道路初始横切面上第i行，第j列第1级搜索点的坐标，本实施例设定道路初始横切面上的第1级搜索点的最上侧行为第0行，依次向下增加，最左侧列为第0列，依次向右增加，如果在第1级搜索点球形邻域范围内存在点云数据，则保留该搜索点，反之则去除，第1级保留下的搜索点称为一级保留搜索点。Among them, N1 and N2 are the interval numbers of search points in the horizontal and vertical directions of the cross section. (x(i,j) ,y(i,j) ,z(i,j) ) are the coordinates of the first-level search point located in the i-th row and j-th column on the initial cross-section of the road. In this embodiment, the road The uppermost row of the first-level search point on the initial cross section is the 0th row, which increases downwards successively, and the leftmost column is the 0th column, which increases successively to the right. If there is For point cloud data, the search point is retained, otherwise, it is removed. The search point retained at the first level is called the first-level reserved search point.
2)第2级搜索点是由一级保留搜索点和一级保留搜索点附近获取的搜索点组成，本发明的一种获取方式为在一级保留搜索点附近获取 4个周围搜索点以及相邻球形邻域所确定4连接搜索点。如图6所示，黑色圆点为道路初始横切面中一级保留搜索点，在图6中单个球形域内，根据一级保留搜索点获取了4个周围的搜索点。4个周围搜索点的公式为：2) The second-level search point is composed of the first-level reserved search point and the search points obtained near the first-level reserved search point. An acquisition method of the present invention is to obtain four surrounding search points and related search points near the first-level reserved search point. 4-connected search points identified by the o-spherical neighborhood. As shown in Figure 6, the black dots are the first-level reserved search points in the initial cross-section of the road. In the single spherical domain in Figure 6, four surrounding search points are obtained based on the first-level reserved search points. The formula for the 4 surrounding search points is:
y2(i,j)＝kid*(x2(i,j)-xi)+y1(k) (13)y2(i,j) =kid *(x2(i,j)-xi )+y1 (k) (13)
(x2(i,j),y2(i,j),z2(i,j))是位于道路初始横切面每个一级球形搜索域上第i行，第j列的第2级搜索点的坐标，本发明的一种实施例设定4 个周围搜索点为两行两列分布，最上侧行为第0行，最下侧行为第1 行，最左侧列为第0列，最右侧列为第1列，但是不仅仅限于这种设定方式；(x1(k),y1(k),z1(k))为一级保留搜索点中的第k个点。在图6中，每个大的球形域中包含4个小的球形域，每一个方形点代表一个小的球形搜索域的球心，半径S为r2，为了防止道路初始横切面出现搜索漏洞，于是在道路初始横切面中相邻球形域的中心范围内再加入了4 个搜索点，即为虚线椭圆所包围的4个搜索点，称之为4个连接搜索点，为了区别一级保留搜索点，这里将4个连接搜索点设置成正方形显示，计算公式为：(x2(i,j) ,y2(i,j) ,z2(i,j) ) is the second level in the i-th row and j-th column of each first-level spherical search domain on the initial cross-section of the road For the coordinates of the search points, an embodiment of the present invention sets the 4 surrounding search points to be distributed in two rows and two columns, the uppermost row is the 0th row, the lowermost row is the 1st row, and the leftmost column is the 0th column. The rightmost column is the first column, but it is not limited to this setting method; (x1 (k), y1 (k), z1 (k)) is the kth point in the first-level reserved search point . In Figure 6, each large spherical domain contains 4 small spherical domains, each square point represents the center of a small spherical search domain, and the radius S is r2 , in order to prevent search holes in the initial cross-section of the road , so 4 search points are added within the center range of the adjacent spherical domain in the initial cross-section of the road, that is, the 4 search points surrounded by the dotted ellipse, which are called 4 connection search points, in order to distinguish the first-level reserved Search point, here the 4 connection search points are set to be displayed in a square, the calculation formula is:
xt1＝λ1*(x1(k+1)+x1(k)) (15)xt1 =λ1 *(x1 (k+1)+x1 (k)) (15)
xt2＝λ2*(x1(k+1)-x1(k))+x1(k) (16)xt2 =λ2 *(x1 (k+1)-x1 (k))+x1 (k) (16)
xt3＝λ3*(x1(k+1)-x1(k))+x1(k) (17)xt3 =λ3 *(x1 (k+1)-x1 (k))+x1 (k) (17)
yt1＝λ1*(y1(k+1)-y1(k)) (18)yt1 =λ1 *(y1 (k+1)-y1 (k)) (18)
yt2＝λ2*(y1(k+1)-y1(k))+y1(k) (19)yt2 =λ2 *(y1 (k+1)-y1 (k))+y1 (k) (19)
yt3＝λ3*(y1(k+1)-y1(k))+y1(k) (20)yt3 =λ3 *(y1 (k+1)-y1 (k))+y1 (k) (20)
zt1＝z1(k)+r2 (21)zt1 =z1 (k)+r2 (21)
zt2＝z1(k)-r2 (22)zt2 =z1 (k)-r2 (22)
zt3＝z1(k) (23)zt3 = z1 (k) (23)
4个连接搜索点坐标分别为(xt1,yt1,zt1)、(xt1,yt1,zt2)、(xt2,yt2,zt3) 和(xt3,yt3,zt3)。本发明实施例中使用的第2级搜索半径为相邻两个搜索点之间距离的0.6倍，但是该距离可以根据具体情况进行调节，如果第2级搜索点的球形邻域内存在点云数据，则保留该搜索点，反之则去除，第2级中保留的搜索点称为二级保留搜索点，本文中λ1、λ1和λ3的值分别为0.5,0.333,0.667。The coordinates of the four connection search points are (xt1 ,yt1 ,zt1 ), (xt1 ,yt1 ,zt2 ), (xt2 ,yt2 ,zt3 ) and (xt3 ,yt3 ,zt3 ). The second-level search radius used in the embodiment of the present invention is 0.6 times the distance between two adjacent search points, but the distance can be adjusted according to specific circumstances, if there is point cloud data in the spherical neighborhood of the second-level search point , the search point is retained, otherwise it is removed. The search point retained in the second level is called the second-level reserved search point. In this paper, the values of λ1 , λ1 and λ3 are 0.5, 0.333, and 0.667, respectively.
至此，获取了道路初始横切面上空含有非空邻域的搜索点集。So far, the search point set with non-empty neighbors above the initial cross-section of the road has been obtained.
S204，将非空邻域内点云向道路横切面投影。S204. Project the point cloud in the non-empty neighborhood to the cross-section of the road.
获取道路初始横切面上含有非空邻域的搜索点集，然后选取这些搜索点球形邻域中点云向道路初始横切面上进行投影，这类点云在初始道路横切面上的投影点，形成最终的道路横切面。Obtain a set of search points with non-empty neighborhoods on the initial cross-section of the road, and then select the point cloud in the spherical neighborhood of these search points to project on the initial cross-section of the road. The projection points of such point clouds on the initial road cross-section, Form the final road cross section.
道路初始横切面的平面方程为：The plane equation of the initial cross section of the road is:
ax+by+cz+d＝0 (24)ax+by+cz+d=0 (24)
若(xi,yi,zi)为道路横切面上的一个轨迹点，则If (xi , yi ,zi ) is a track point on the cross section of the road, then
d＝-(axi+byi+czi) (25)d＝-(axi +byi +czi ) (25)
a,b和c的值在本文中分别是xid,yid和cid，The values of a, b and c are respectively xid , yid and cid in this paper,
每一个原始三维点向横切面上的投影公式为：The projection formula of each original 3D point to the cross section is:
其中(xsi,ysi,zsi)是球形邻域中的点云坐标，(xp,yp,zp)为道路初始横切面上的投影点。Where (xsi , ysi , zsi ) are point cloud coordinates in the spherical neighborhood, and (xp , yp , zp ) are projected points on the initial cross-section of the road.
再次参考图1，本发明的方法还包括S30。在S30，通过单线点云提取道路横切面的轮廓。如图7所示，步骤S30包括S301-S303。Referring to FIG. 1 again, the method of the present invention further includes S30. At S30, the outline of the cross-section of the road is extracted from the single-line point cloud. As shown in FIG. 7, step S30 includes S301-S303.
S301，获取单线点云S301, acquiring a single-line point cloud
单线点云上的三维点依次排列形成一条直线，其长度由可以根据道路的宽度设定，位于道路横切面共面且位于道路底面下方，单线点云中相邻点的间隔为d m，本发明实施例中左右拓展长度分别为7m,d 的设定值0.02，但是相邻点的间隔和左右拓展长度可以根据道路实际情况赋值，单线点云如图8所示,即其中的由点组成的灰色线段。The three-dimensional points on the single-line point cloud are arranged in sequence to form a straight line, the length of which can be set according to the width of the road, which is located on the same plane as the cross-section of the road and below the bottom surface of the road, and the distance between adjacent points in the single-line point cloud is d m. The present invention In the embodiment, the left and right extension lengths are respectively 7m, and the set value of d is 0.02, but the interval between adjacent points and the left and right extension lengths can be assigned according to the actual situation of the road. gray lines.
S302，获取非空邻域单线点S302. Obtain a single-line point in a non-empty neighborhood
如图8所示，单线点云向道路横切面逐渐靠近，在上升过程中，利用kd-tree近似最邻近邻域搜索算法来搜索单线点云中每一个单线点云中的点的球形邻域中是否存在道路环境中的原始三维点,若存在，则这些点就停止搜索，执行获取道路横切面轮廓点操作，单线点云上剩余点的继续搜索。在一定的迭代次数内，获取所有具有非空邻域的单线点。As shown in Figure 8, the single-line point cloud is gradually approaching the cross-section of the road. During the ascending process, the kd-tree approximate nearest neighbor search algorithm is used to search for the spherical neighborhood of each point in the single-line point cloud Whether there are original 3D points in the road environment, if so, stop searching for these points, execute the operation of obtaining road cross-section contour points, and continue searching for the remaining points on the single-line point cloud. Within a certain number of iterations, get all single-line points with non-empty neighbors.
S303，获取道路横切面轮廓点S303, obtaining road cross-section contour points
获取道路横切面轮廓点操作是给单线点云中具有非空邻域的单线点集分配新的位置，形成道路横切面的轮廓，通过该操作使单线点云中的点更加的贴近道路横切面上的点，保证了获取的道路轮廓的完整性和平滑性。新的位置坐标(xpos,ypos,zpos)计算公式如下：The operation of obtaining the contour points of the road cross-section is to assign a new position to the single-line point set with a non-empty neighborhood in the single-line point cloud to form the contour of the road cross-section. Through this operation, the points in the single-line point cloud are closer to the road cross-section The points above ensure the integrity and smoothness of the acquired road profile. The calculation formula of the new position coordinates (xpos , ypos , zpos ) is as follows:
其中n为具有非空邻域单线点的球形邻域中三维点的个数， (xsi,ysi,zsi)为具有非空邻域单线点的球形邻域中原始三维点的坐标，如图8所示描述了单线点云逐渐靠近道路轮廓的过程，图中红色的点为道路边界点，最终的单线点云完整的附着在道路的表面，得到道路边界的轮廓。where n is the number of three-dimensional points in the spherical neighborhood with non-empty neighborhood single-line points, (xsi , ysi , zsi ) are the coordinates of the original three-dimensional points in the spherical neighborhood with non-empty neighborhood single-line points, As shown in Figure 8, the process of the single-line point cloud gradually approaching the road outline is described. The red points in the figure are the road boundary points. The final single-line point cloud is completely attached to the road surface, and the outline of the road boundary is obtained.
再次参考图1，本发明的方法还包括S40。在S40，根据道路边界的特征，通过窗口法检测道路边界点。Referring to FIG. 1 again, the method of the present invention further includes S40. At S40, the road boundary points are detected by the window method according to the characteristics of the road boundary.
道路边界点的特征在很多的方面与道路平面点的特征有差别。首先道路边界区域的高度差更大，其次道路高程在垂直于道路边界的方向出现剧烈变化，最后道路边界点是平滑延伸分布的。这三种特征总结为道路高程跳变特征，斜率变化和道路边界点的分布特征。本发明基于这三种特征利用窗口方法对在道路轮廓中的道路边界点进行检测。首先指定两个窗口，从道路轮廓的中点开始向两侧搜索。具体的道路边界点检测方法如下：The characteristics of road boundary points are different from those of road plane points in many aspects. Firstly, the height difference in the road boundary area is larger, secondly, the road elevation changes drastically in the direction perpendicular to the road boundary, and finally, the road boundary points are distributed smoothly. These three characteristics are summarized as road elevation jump characteristics, slope changes and distribution characteristics of road boundary points. The present invention uses the window method to detect the road boundary points in the road outline based on these three features. First specify two windows, and search from the midpoint of the road profile to both sides. The specific road boundary point detection method is as follows:
1)高程跳变1) Elevation jump
如图9A所示，窗口1中包含道路边界的高程跳变。在图9B中道路边界的高程跳变存在两个窗口中。本发明将两个窗口融合成一个大的窗口，在大窗口中进行高程跳变检测。在该过程中，首先确定高程跳变存在的位置，高程跳变计算公式为：As shown in FIG. 9A , window 1 includes elevation jumps of road boundaries. In Fig. 9B, the elevation jump of the road boundary exists in two windows. The invention fuses two windows into one large window, and performs height jump detection in the large window. In this process, first determine the position where the elevation jump exists, and the calculation formula of the elevation jump is:
其中Δz1是窗口1存在的最大高程差，Δz2为窗口2存在的最大高程差，Δz12为窗口1与窗口2之间存在的最大高程差，道路边界的高度设定范围为Hc1和Hc2，两者的取值可以根据道路的实际情概况进行调节。如果Hc1<Δz1<Hc2，则证明道路边界点位于窗口1中，然后在窗口1中检测道路边界点。如果Hc1<Δz2<Hc2，证明道路边界位于窗口2 中。如果Hc1<Δz12<Hc2则道路边界点需要两个窗口共同确定。本文实施例中的Hc1为0.1m,Hc2为0.25m。Where Δz1 is the maximum elevation difference in window 1, Δz2 is the maximum elevation difference in window 2, Δz12 is the maximum elevation difference between window 1 and window 2, and the height setting range of the road boundary is Hc1 and Hc2 , the values of the two can be adjusted according to the actual situation of the road. If Hc1 <Δz1 <Hc2 , it is proved that the road boundary point is located in window 1, and then the road boundary point is detected in window 1. If Hc1 <Δz2 <Hc2 , it is proved that the road boundary is in window 2 . If Hc1 <Δz12 <Hc2 , the road boundary points need to be determined jointly by two windows.Hc1 in the examples herein is 0.1m, and Hc2is 0.25m.
如果高程跳变在单个窗口中，则在该窗口中检测道路边界边界点。单窗口中道路边界点的检测公式为：If the elevation jump is in a single window, the road boundary boundary points are detected in that window. The detection formula of road boundary points in a single window is:
Δz＝|zw(i)-zw(i+1)| (29)Δz＝|zw (i)-zw (i+1)| (29)
如果Hc1<Δz<Hc2，则可以确定道路横切面轮廓中第i个三维点满足道路边界点的高程跳变情况。如果道路边界点分布在两个窗口内，融合两个流体窗口为一个窗口，检测方式与单一窗口检测方式相同。 zw(i)为窗口中第i个道路横切面轮廓点的z值，zw(i+1)为窗口中第i+1 个道路横切面轮廓点的z值。If Hc1 <Δz<Hc2 , it can be determined that the i-th three-dimensional point in the cross-section profile of the road satisfies the elevation jump of the road boundary point. If the road boundary points are distributed in two windows, the two fluid windows are fused into one window, and the detection method is the same as the single window detection method. zw (i) is the z value of the i-th road cross-section contour point in the window, and zw (i+1) is the z-value of the i+1 road cross-section contour point in the window.
2)斜率2) Slope
由于道路边界垂直于道路平面，在道路边界凸起的地方斜率会出现剧烈变化，本文通过道路边界点和其相邻点存在较大斜率值。因此可以相邻两个道路轮廓点的斜率值来判断道路边界点的位置。斜率s 的计算公式为：Since the road boundary is perpendicular to the road plane, the slope of the road boundary will change drastically. In this paper, the road boundary point and its adjacent points have a large slope value. Therefore, the position of the road boundary point can be judged by the slope value of two adjacent road contour points. The formula for calculating the slope s is:
道路边界点的斜率阈值为sT,如果s>sT，则坐标为 (xw(i),yw(i),zw(i))的道路横切面轮廓点满足道路边界点斜率的特征。The slope threshold of the road boundary point is sT , if s>sT , the road cross-section contour points with coordinates (xw (i), yw (i), zw (i)) meet the slope threshold of the road boundary point feature.
3)道路边界点的分布3) Distribution of road boundary points
道路边界具有连续延伸的特征，相邻道路横切面上同侧道路边界点在道路延伸的方向会偏移较大的距离，在其它两个方向的距离偏移较小，如图3所示：该实施例中道路边界点的分布特征为两个相邻的边界点在在y方向上存在较大的距离，在x和z方向存在较小的距离，根据道路边界点的分布特征来判断检测到的道路边界点是否正确，剔除错误的边界点。The road boundary has the characteristics of continuous extension. The road boundary points on the same side on the adjacent road cross-section will deviate a large distance in the direction of road extension, and the distance deviation in the other two directions is small, as shown in Figure 3: The distribution feature of the road boundary points in this embodiment is that two adjacent boundary points have a larger distance in the y direction, and a smaller distance in the x and z directions, and the detection is judged according to the distribution characteristics of the road boundary points Whether the road boundary points arrived are correct, and the wrong boundary points are eliminated.
相邻边界点y方向的距离Dy计算公式为：The formula for calculating the distance Dy in the y direction of adjacent boundary points is:
Dy＝|yi-yi+1| (31)Dy =|yi -yi+1 | (31)
相邻边界点xz方向的距离Dxz计算公式为：The formula for calculating the distance Dxz in the xz direction of adjacent boundary points is:
如果Dy>DT1且Dxz<DT2，检测到的边界点满足道路边界连续延伸特性。其中(xi,yi,zi)和(xi+1,yi+1,zi+1)分别为相邻道路边界点的坐标，DT1和DT2分别是相邻边界点在y方向和xz方向的距离阈值，DT1和DT2的设定和相邻横切面之间的距离相关。本文中横切面之间距离为0.5m, 设定DT1为0.45m,DT2为0.05m。If Dy >DT1 and Dxz <DT2 , the detected boundary points satisfy the road boundary continuous extension characteristic. Where (xi , yi , zi ) and (xi+1 , yi+1 , zi+1 ) are the coordinates of adjacent road boundary points respectively, DT1 and DT2 are the adjacent boundary points in The distance thresholds in the y direction and xz direction, the settings of DT1 and DT2 are related to the distance between adjacent cross-sections. In this paper, the distance between cross-sections is 0.5m,DT1 is set to 0.45m, andDT2 is set to 0.05m.
再次参考图1，本发明的方法还包括S50。在S50，拟合道路边界点获取完整的道路边界线。Referring to FIG. 1 again, the method of the present invention further includes S50. At S50, the road boundary points are fitted to obtain a complete road boundary line.
在实际的道路情况下，由于车辆或者行人的遮挡导致道路边界在数据中出现空洞的现象，即被遮挡的道路边界在数据上未被记录，或者提出的算法中存在少量的外点，并且提取的道路边界点之间存在着较大距离，所以我们有必要对道路边界点进行优化，然后对优化后的点集进行道路边界线的拟合。拟合的方法有多种，最小二乘法， Ransanc法以及二次曲线拟合法等。本实施例中通过Ransac方法进行道路边界点的优化和道路边界线参数的获取。本实施例假设道路边界的形状为直线型，为了进一步的提高道路边界检测的正确率和精度采取Ransac算法对道路边界点进行拟合。本实施例中Ransac算法能鲁棒的估计直线模型参数，它能从包含局外点的数据集中估计出高精度三维道路边界直线的参数，通过直线参数和直线的起始位置绘制出完整的道路边界线。In the actual road situation, due to the occlusion of vehicles or pedestrians, there are holes in the road boundary in the data, that is, the occluded road boundary is not recorded in the data, or there are a small number of outliers in the proposed algorithm, and the extraction There is a large distance between the road boundary points, so it is necessary for us to optimize the road boundary points, and then fit the road boundary line to the optimized point set. There are many fitting methods, such as least squares method, Ransanc method and quadratic curve fitting method. In this embodiment, the optimization of road boundary points and the acquisition of road boundary line parameters are performed by Ransac method. In this embodiment, it is assumed that the shape of the road boundary is linear, and in order to further improve the accuracy and accuracy of road boundary detection, the Ransac algorithm is used to fit the road boundary points. In this embodiment, the Ransac algorithm can robustly estimate the parameters of the straight line model. It can estimate the parameters of the high-precision three-dimensional road boundary straight line from the data set containing outliers, and draw a complete road through the straight line parameters and the starting position of the straight line. borderline.
本发明的方法进行了实验，实验中采集的是北京市三环路段上的三维数据，根据道路场景复杂程度挑选了3种不同类型的数据用于检测算法有效性，分别是城市环境中无遮挡的道路边界，被绿化带遮挡的道路边界和被车辆遮挡的道路边界，三种类型的数据如下图所示。图10白色边框中是无绿化带遮挡的道路边界，该道路边界的存在平整的上边界平面和下边界平面，即规则的道路边界模型。图11白色边框中道路边界的上平面存在绿色的杂点，实际上是杂草覆盖到了道路边界上平面，这样造成了道路边界的整体形态发生了无规则的变化，但是道路边界的地面特征仍是平滑的。图12白色框中是汽车遮挡道路边界的情形，白色框中由于道路边界被遮挡而出现了空洞现象。The method of the present invention has been tested. What was collected in the experiment was the three-dimensional data on the third ring road section of Beijing. According to the complexity of the road scene, three different types of data were selected for testing the effectiveness of the algorithm. The road boundary of , the road boundary blocked by green belts and the road boundary blocked by vehicles, the three types of data are shown in the figure below. The white frame in Figure 10 is the road boundary without green belts, and the road boundary has a flat upper boundary plane and a lower boundary plane, that is, a regular road boundary model. In Figure 11, there are green noise spots on the upper plane of the road boundary in the white frame. In fact, weeds cover the upper plane of the road boundary, which causes irregular changes in the overall shape of the road boundary, but the ground features of the road boundary remain the same. is smooth. The white box in Figure 12 is the situation where the road boundary is blocked by the car, and there is a hollow phenomenon in the white box due to the occlusion of the road boundary.
如图13A所示，道路边界完整度高，没有道路杂草的覆盖影响，道路边界的检测结果如图13B所示，其中白色的直线为检测到的道路边界，获取的道路边界准确的与实际的道路边界相重合。As shown in Figure 13A, the integrity of the road boundary is high, and there is no influence of road weed coverage. The detection result of the road boundary is shown in Figure 13B, where the white straight line is the detected road boundary, and the obtained road boundary is accurate and actual coincide with the road boundaries.
如图14A所示，左道路边界完整度高，右道路边界出现杂草遮挡，如图中白框所圈区域。虽然右边道路边界存在大量的杂草点，导致道路边界上平面形态遭到严重的破坏，但是道路边界的下平面形态仍然保存完整。本发明针对这一特点进行单线点云从下到上检测道路的整体轮廓，避免了道路表面出现遮挡时对道路边界的影响。如图 14B所示，通过本发明方法，左右道路边界同时被完整的检测出来，如图中白线所示，两条检测到的边界线与实际道路边界相融合证明了本发明的在检测道路边界存在绿化带遮挡情况下的有效性。As shown in Figure 14A, the integrity of the left road boundary is high, and the right road boundary is covered by weeds, such as the area circled by the white box in the figure. Although there are a large number of weed points on the right road boundary, resulting in serious damage to the plane shape of the road boundary, the lower plane shape of the road boundary is still intact. Aiming at this characteristic, the present invention detects the overall outline of the road from the bottom to the top with a single-line point cloud, and avoids the influence on the road boundary when the road surface is blocked. As shown in Figure 14B, through the method of the present invention, the left and right road boundaries are completely detected at the same time. The validity of the boundary is covered by green belts.
图15A-15B是在道路边界存在车辆遮挡的情况下的道路边界检测的结果。道路右边界在汽车遮挡的情况下，激光器无法获取道路边界的信息，在图15A中出现空洞的现象，白框包围处，本发明在借助Ransac算法和空洞前后存在较好的道路边界点的情况下，拟合出了一条完整的道路边界，如图15B与实际边界的位移较小，取得较好的效果。15A-15B are the results of road boundary detection in the case of vehicle occlusion at the road boundary. When the right boundary of the road is blocked by the car, the laser cannot obtain the information of the road boundary, and a hole appears in Figure 15A, surrounded by a white frame, the present invention has better road boundary points before and after the use of the Ransac algorithm and the hole Next, a complete road boundary is fitted, as shown in Figure 15B, the displacement of the actual boundary is small, and a good effect is achieved.
以上所述的实施例，只是本发明较优选的具体实施方式，本领域的技术人员在本发明技术方案范围内进行的通常变化和替换都应包含在本发明的保护范围内。The above-described embodiments are only preferred specific implementations of the present invention, and ordinary changes and replacements performed by those skilled in the art within the scope of the technical solution of the present invention should be included in the protection scope of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710376969.2ACN107169464B (en) | 2017-05-25 | 2017-05-25 | A road boundary detection method based on laser point cloud |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710376969.2ACN107169464B (en) | 2017-05-25 | 2017-05-25 | A road boundary detection method based on laser point cloud |
| Publication Number | Publication Date |
|---|---|
| CN107169464Atrue CN107169464A (en) | 2017-09-15 |
| CN107169464B CN107169464B (en) | 2019-04-09 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710376969.2AActiveCN107169464B (en) | 2017-05-25 | 2017-05-25 | A road boundary detection method based on laser point cloud |
| Country | Link |
|---|---|
| CN (1) | CN107169464B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108074232A (en)* | 2017-12-18 | 2018-05-25 | 辽宁工程技术大学 | A kind of airborne LIDAR based on volume elements segmentation builds object detecting method |
| CN108319895A (en)* | 2017-12-29 | 2018-07-24 | 百度在线网络技术（北京）有限公司 | The method and apparatus at the crossing in electronic map for identification |
| CN108829951A (en)* | 2018-05-31 | 2018-11-16 | 北京中油瑞飞信息技术有限责任公司 | Reservoir model boundary determining method and device |
| CN108931786A (en)* | 2018-05-17 | 2018-12-04 | 北京智行者科技有限公司 | Curb detection device and method |
| CN108957432A (en)* | 2018-04-09 | 2018-12-07 | 深圳清创新科技有限公司 | Curb detection method, device, computer equipment and storage medium |
| CN109657628A (en)* | 2018-12-24 | 2019-04-19 | 驭势（上海）汽车科技有限公司 | It is a kind of for determining the method and apparatus in the travelable region of vehicle |
| CN109738910A (en)* | 2019-01-28 | 2019-05-10 | 重庆邮电大学 | A road edge detection method based on 3D lidar |
| CN109766404A (en)* | 2019-02-12 | 2019-05-17 | 湖北亿咖通科技有限公司 | Points cloud processing method, apparatus and computer readable storage medium |
| CN110008921A (en)* | 2019-04-12 | 2019-07-12 | 北京百度网讯科技有限公司 | A method, device, electronic device and storage medium for generating road boundary |
| CN110068834A (en)* | 2018-01-24 | 2019-07-30 | 北京京东尚科信息技术有限公司 | A kind of curb detection method and device |
| CN110222605A (en)* | 2019-05-24 | 2019-09-10 | 深兰科技（上海）有限公司 | A kind of obstacle detection method and equipment |
| CN110246092A (en)* | 2019-05-02 | 2019-09-17 | 江西理工大学 | A kind of three-dimensional laser point cloud denoising method for taking neighborhood point mean distance and slope into account |
| CN110378173A (en)* | 2018-07-10 | 2019-10-25 | 北京京东尚科信息技术有限公司 | A kind of method and apparatus of determining lane boundary line |
| CN110618413A (en)* | 2018-06-19 | 2019-12-27 | 北京京东尚科信息技术有限公司 | Passable area detection method and device based on multi-line laser radar |
| CN110674705A (en)* | 2019-09-05 | 2020-01-10 | 北京智行者科技有限公司 | Small-sized obstacle detection method and device based on multi-line laser radar |
| CN110696826A (en)* | 2019-10-09 | 2020-01-17 | 北京百度网讯科技有限公司 | Method and apparatus for controlling a vehicle |
| CN110967024A (en)* | 2019-12-23 | 2020-04-07 | 苏州智加科技有限公司 | Method, device, equipment and storage medium for detecting travelable area |
| CN111007531A (en)* | 2019-12-24 | 2020-04-14 | 电子科技大学 | A road edge detection method based on laser point cloud data |
| CN111076734A (en)* | 2019-12-12 | 2020-04-28 | 湖南大学 | A high-precision map construction method for unstructured roads in closed areas |
| CN111144211A (en)* | 2019-08-28 | 2020-05-12 | 华为技术有限公司 | Point cloud display method and device |
| WO2020093966A1 (en)* | 2018-11-09 | 2020-05-14 | 阿里巴巴集团控股有限公司 | Positioning data generation method, apparatus, and electronic device |
| CN111273305A (en)* | 2020-02-18 | 2020-06-12 | 中国科学院合肥物质科学研究院 | Multi-sensor fusion road extraction and indexing method based on global and local grid maps |
| CN111325138A (en)* | 2020-02-18 | 2020-06-23 | 中国科学院合肥物质科学研究院 | A real-time detection method of road boundary based on local bump feature of point cloud |
| CN111323026A (en)* | 2018-12-17 | 2020-06-23 | 兰州大学 | Ground filtering method based on high-precision point cloud map |
| US10901421B2 (en) | 2018-06-26 | 2021-01-26 | Neusoft Reach Automotive Technology (Shanghai) Co., Ltd. | Method and device for detecting road boundary |
| CN112384962A (en)* | 2018-07-02 | 2021-02-19 | 日产自动车株式会社 | Driving assistance method and driving assistance device |
| CN113378800A (en)* | 2021-07-27 | 2021-09-10 | 武汉市测绘研究院 | Automatic classification and vectorization method for road sign lines based on vehicle-mounted three-dimensional point cloud |
| CN113379776A (en)* | 2021-05-14 | 2021-09-10 | 北京踏歌智行科技有限公司 | Road boundary detection method |
| CN113447953A (en)* | 2021-06-29 | 2021-09-28 | 山东高速建设管理集团有限公司 | Background filtering method based on road traffic point cloud data |
| CN113610883A (en)* | 2021-04-30 | 2021-11-05 | 新驱动重庆智能汽车有限公司 | Point cloud processing system and method, computer device, and storage medium |
| CN113778081A (en)* | 2021-08-19 | 2021-12-10 | 中国农业科学院农业资源与农业区划研究所 | Orchard path identification method and robot based on laser radar and vision |
| CN113807193A (en)* | 2021-08-23 | 2021-12-17 | 武汉中海庭数据技术有限公司 | Method and system for automatically extracting virtual line segments of traffic roads in laser point cloud |
| CN113836978A (en)* | 2020-06-24 | 2021-12-24 | 富士通株式会社 | Road area determination device and method, electronic equipment |
| CN114118120A (en)* | 2021-03-04 | 2022-03-01 | 贵州京邦达供应链科技有限公司 | Data processing method, device, electronic device and storage medium |
| WO2022042359A1 (en)* | 2020-08-26 | 2022-03-03 | 深圳市杉川机器人有限公司 | Working area map establishing method and self-moving device |
| CN115702450A (en)* | 2020-06-23 | 2023-02-14 | 株式会社电装 | Vehicle position estimation device and travel position estimation method |
| CN116051753A (en)* | 2023-02-16 | 2023-05-02 | 北京瑞医博科技有限公司 | Three-dimensional modeling and three-dimensional printing method and device, electronic equipment and storage medium |
| CN116164775A (en)* | 2022-11-30 | 2023-05-26 | 武汉中海庭数据技术有限公司 | A method and device for automatic detection of road surface bumps and convexities on OpenDRIVE |
| CN116563821A (en)* | 2023-05-05 | 2023-08-08 | 阿里巴巴达摩院(杭州)科技有限公司 | Region detection and visual perception model training method, electronic equipment and storage medium |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0728975A (en)* | 1993-06-25 | 1995-01-31 | Fujitsu Ltd | Road environment recognition device |
| CN1838174A (en)* | 2005-03-22 | 2006-09-27 | 日产自动车株式会社 | Detecting device and method to detect an object based on a road boundary |
| US20090052742A1 (en)* | 2007-08-24 | 2009-02-26 | Kabushiki Kaisha Toshiba | Image processing apparatus and method thereof |
| CN101901343A (en)* | 2010-07-20 | 2010-12-01 | 同济大学 | Road Extraction Method Based on Stereo Constraint from Remote Sensing Image |
| CN102270301A (en)* | 2011-06-07 | 2011-12-07 | 南京理工大学 | Method for detecting unstructured road boundary by combining support vector machine (SVM) and laser radar |
| CN104272363A (en)* | 2012-04-27 | 2015-01-07 | 株式会社电装 | Boundary-line-detecting device, boundary-line-deviation-detecting device, and boundary-line-detecting program |
| CN104850834A (en)* | 2015-05-11 | 2015-08-19 | 中国科学院合肥物质科学研究院 | Road boundary detection method based on three-dimensional laser radar |
| CN105404844A (en)* | 2014-09-12 | 2016-03-16 | 广州汽车集团股份有限公司 | Road boundary detection method based on multi-line laser radar |
| CN106503678A (en)* | 2016-10-27 | 2017-03-15 | 厦门大学 | Roadmarking automatic detection and sorting technique based on mobile laser scanning point cloud |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0728975A (en)* | 1993-06-25 | 1995-01-31 | Fujitsu Ltd | Road environment recognition device |
| CN1838174A (en)* | 2005-03-22 | 2006-09-27 | 日产自动车株式会社 | Detecting device and method to detect an object based on a road boundary |
| US20090052742A1 (en)* | 2007-08-24 | 2009-02-26 | Kabushiki Kaisha Toshiba | Image processing apparatus and method thereof |
| CN101901343A (en)* | 2010-07-20 | 2010-12-01 | 同济大学 | Road Extraction Method Based on Stereo Constraint from Remote Sensing Image |
| CN102270301A (en)* | 2011-06-07 | 2011-12-07 | 南京理工大学 | Method for detecting unstructured road boundary by combining support vector machine (SVM) and laser radar |
| CN104272363A (en)* | 2012-04-27 | 2015-01-07 | 株式会社电装 | Boundary-line-detecting device, boundary-line-deviation-detecting device, and boundary-line-detecting program |
| CN105404844A (en)* | 2014-09-12 | 2016-03-16 | 广州汽车集团股份有限公司 | Road boundary detection method based on multi-line laser radar |
| CN104850834A (en)* | 2015-05-11 | 2015-08-19 | 中国科学院合肥物质科学研究院 | Road boundary detection method based on three-dimensional laser radar |
| CN106503678A (en)* | 2016-10-27 | 2017-03-15 | 厦门大学 | Roadmarking automatic detection and sorting technique based on mobile laser scanning point cloud |
| Title |
|---|
| BISHENGYANG ET AL: "Semi-automated extraction and delineation of 3D roads of street scene from mobile laser scanning point clouds", 《ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING》* |
| BORJARODRÍGUEZ-CUENCA ET AL: "An approach to detect and delineate street curbs from MLS 3D point cloud data", 《AUTOMATION IN CONSTRUCTION》* |
| SHENG XU ET AL: "Road Curb Extraction From Mobile LiDAR Point Clouds", 《IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING》* |
| SHERIF EL-HALAWANY ET AL: "DETECTION OF ROAD CURB FROM MOBILE TERRESTRIAL LASER SCANNER", 《INTERNATIONAL ARCHIVES OF THE PHOTOGRAMMETRY, REMOTE SENSING AND SPATIAL INFORMATION SCIENCES》* |
| 徐杰等: "车辆自主导航中的道路边界识别算法", 《中国图象图形学报》* |
| 谭宝成等: "基于激光雷达的道路边界检测的研究", 《电子设计工程》* |
| 邹晓亮等: "移动车载激光点云的道路标线自动识别与提取", 《测绘与空间地理信息》* |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108074232B (en)* | 2017-12-18 | 2021-08-24 | 辽宁工程技术大学 | An airborne LIDAR building detection method based on voxel segmentation |
| CN108074232A (en)* | 2017-12-18 | 2018-05-25 | 辽宁工程技术大学 | A kind of airborne LIDAR based on volume elements segmentation builds object detecting method |
| CN108319895A (en)* | 2017-12-29 | 2018-07-24 | 百度在线网络技术（北京）有限公司 | The method and apparatus at the crossing in electronic map for identification |
| CN110068834B (en)* | 2018-01-24 | 2023-04-07 | 北京京东尚科信息技术有限公司 | Road edge detection method and device |
| CN110068834A (en)* | 2018-01-24 | 2019-07-30 | 北京京东尚科信息技术有限公司 | A kind of curb detection method and device |
| CN108957432B (en)* | 2018-04-09 | 2021-01-12 | 深圳一清创新科技有限公司 | Road edge detection method and device, computer equipment and storage medium |
| CN108957432A (en)* | 2018-04-09 | 2018-12-07 | 深圳清创新科技有限公司 | Curb detection method, device, computer equipment and storage medium |
| CN108931786A (en)* | 2018-05-17 | 2018-12-04 | 北京智行者科技有限公司 | Curb detection device and method |
| CN108829951A (en)* | 2018-05-31 | 2018-11-16 | 北京中油瑞飞信息技术有限责任公司 | Reservoir model boundary determining method and device |
| CN110618413A (en)* | 2018-06-19 | 2019-12-27 | 北京京东尚科信息技术有限公司 | Passable area detection method and device based on multi-line laser radar |
| US10901421B2 (en) | 2018-06-26 | 2021-01-26 | Neusoft Reach Automotive Technology (Shanghai) Co., Ltd. | Method and device for detecting road boundary |
| CN112384962B (en)* | 2018-07-02 | 2022-06-21 | 日产自动车株式会社 | Driving assistance method and driving assistance device |
| CN112384962A (en)* | 2018-07-02 | 2021-02-19 | 日产自动车株式会社 | Driving assistance method and driving assistance device |
| CN110378173A (en)* | 2018-07-10 | 2019-10-25 | 北京京东尚科信息技术有限公司 | A kind of method and apparatus of determining lane boundary line |
| WO2020093966A1 (en)* | 2018-11-09 | 2020-05-14 | 阿里巴巴集团控股有限公司 | Positioning data generation method, apparatus, and electronic device |
| CN111323026B (en)* | 2018-12-17 | 2023-07-07 | 兰州大学 | A Ground Filtering Method Based on High Precision Point Cloud Map |
| CN111323026A (en)* | 2018-12-17 | 2020-06-23 | 兰州大学 | Ground filtering method based on high-precision point cloud map |
| CN109657628A (en)* | 2018-12-24 | 2019-04-19 | 驭势（上海）汽车科技有限公司 | It is a kind of for determining the method and apparatus in the travelable region of vehicle |
| CN109738910A (en)* | 2019-01-28 | 2019-05-10 | 重庆邮电大学 | A road edge detection method based on 3D lidar |
| CN109766404B (en)* | 2019-02-12 | 2020-12-15 | 湖北亿咖通科技有限公司 | Point cloud processing method and device and computer readable storage medium |
| CN109766404A (en)* | 2019-02-12 | 2019-05-17 | 湖北亿咖通科技有限公司 | Points cloud processing method, apparatus and computer readable storage medium |
| CN110008921A (en)* | 2019-04-12 | 2019-07-12 | 北京百度网讯科技有限公司 | A method, device, electronic device and storage medium for generating road boundary |
| CN110008921B (en)* | 2019-04-12 | 2021-12-28 | 北京百度网讯科技有限公司 | Road boundary generation method and device, electronic equipment and storage medium |
| CN110246092B (en)* | 2019-05-02 | 2022-09-02 | 江西理工大学 | Three-dimensional laser point cloud denoising method considering neighborhood point mean distance and slope |
| CN110246092A (en)* | 2019-05-02 | 2019-09-17 | 江西理工大学 | A kind of three-dimensional laser point cloud denoising method for taking neighborhood point mean distance and slope into account |
| CN110222605A (en)* | 2019-05-24 | 2019-09-10 | 深兰科技（上海）有限公司 | A kind of obstacle detection method and equipment |
| CN110222605B (en)* | 2019-05-24 | 2021-11-19 | 深兰科技（上海）有限公司 | Obstacle detection method and equipment |
| CN111144211A (en)* | 2019-08-28 | 2020-05-12 | 华为技术有限公司 | Point cloud display method and device |
| CN111144211B (en)* | 2019-08-28 | 2023-09-12 | 华为技术有限公司 | Point cloud display method and device |
| CN110674705A (en)* | 2019-09-05 | 2020-01-10 | 北京智行者科技有限公司 | Small-sized obstacle detection method and device based on multi-line laser radar |
| CN110696826A (en)* | 2019-10-09 | 2020-01-17 | 北京百度网讯科技有限公司 | Method and apparatus for controlling a vehicle |
| CN110696826B (en)* | 2019-10-09 | 2022-04-01 | 北京百度网讯科技有限公司 | Method and device for controlling a vehicle |
| CN111076734B (en)* | 2019-12-12 | 2021-07-23 | 湖南大学 | A high-precision map construction method for unstructured roads in closed areas |
| CN111076734A (en)* | 2019-12-12 | 2020-04-28 | 湖南大学 | A high-precision map construction method for unstructured roads in closed areas |
| CN110967024A (en)* | 2019-12-23 | 2020-04-07 | 苏州智加科技有限公司 | Method, device, equipment and storage medium for detecting travelable area |
| WO2021128777A1 (en)* | 2019-12-23 | 2021-07-01 | Suzhou Zhijia Science & Technologies Co., Ltd. | Method, apparatus, device, and storage medium for detecting travelable region |
| CN111007531A (en)* | 2019-12-24 | 2020-04-14 | 电子科技大学 | A road edge detection method based on laser point cloud data |
| CN111325138A (en)* | 2020-02-18 | 2020-06-23 | 中国科学院合肥物质科学研究院 | A real-time detection method of road boundary based on local bump feature of point cloud |
| CN111325138B (en)* | 2020-02-18 | 2023-04-07 | 中国科学院合肥物质科学研究院 | Road boundary real-time detection method based on point cloud local concave-convex characteristics |
| CN111273305A (en)* | 2020-02-18 | 2020-06-12 | 中国科学院合肥物质科学研究院 | Multi-sensor fusion road extraction and indexing method based on global and local grid maps |
| CN115702450A (en)* | 2020-06-23 | 2023-02-14 | 株式会社电装 | Vehicle position estimation device and travel position estimation method |
| CN115702450B (en)* | 2020-06-23 | 2025-08-26 | 株式会社电装 | Vehicle position estimation device and driving position estimation method |
| CN113836978A (en)* | 2020-06-24 | 2021-12-24 | 富士通株式会社 | Road area determination device and method, electronic equipment |
| WO2022042359A1 (en)* | 2020-08-26 | 2022-03-03 | 深圳市杉川机器人有限公司 | Working area map establishing method and self-moving device |
| CN114118120A (en)* | 2021-03-04 | 2022-03-01 | 贵州京邦达供应链科技有限公司 | Data processing method, device, electronic device and storage medium |
| CN113610883A (en)* | 2021-04-30 | 2021-11-05 | 新驱动重庆智能汽车有限公司 | Point cloud processing system and method, computer device, and storage medium |
| CN113379776A (en)* | 2021-05-14 | 2021-09-10 | 北京踏歌智行科技有限公司 | Road boundary detection method |
| CN113447953A (en)* | 2021-06-29 | 2021-09-28 | 山东高速建设管理集团有限公司 | Background filtering method based on road traffic point cloud data |
| CN113447953B (en)* | 2021-06-29 | 2022-08-26 | 山东高速建设管理集团有限公司 | Background filtering method based on road traffic point cloud data |
| CN113378800A (en)* | 2021-07-27 | 2021-09-10 | 武汉市测绘研究院 | Automatic classification and vectorization method for road sign lines based on vehicle-mounted three-dimensional point cloud |
| CN113378800B (en)* | 2021-07-27 | 2021-11-09 | 武汉市测绘研究院 | Automatic classification and vectorization method for road sign lines based on vehicle-mounted three-dimensional point cloud |
| CN113778081B (en)* | 2021-08-19 | 2022-07-22 | 中国农业科学院农业资源与农业区划研究所 | A method and robot for orchard path recognition based on lidar and vision |
| CN113778081A (en)* | 2021-08-19 | 2021-12-10 | 中国农业科学院农业资源与农业区划研究所 | Orchard path identification method and robot based on laser radar and vision |
| CN113807193A (en)* | 2021-08-23 | 2021-12-17 | 武汉中海庭数据技术有限公司 | Method and system for automatically extracting virtual line segments of traffic roads in laser point cloud |
| CN116164775A (en)* | 2022-11-30 | 2023-05-26 | 武汉中海庭数据技术有限公司 | A method and device for automatic detection of road surface bumps and convexities on OpenDRIVE |
| CN116051753A (en)* | 2023-02-16 | 2023-05-02 | 北京瑞医博科技有限公司 | Three-dimensional modeling and three-dimensional printing method and device, electronic equipment and storage medium |
| CN116563821A (en)* | 2023-05-05 | 2023-08-08 | 阿里巴巴达摩院(杭州)科技有限公司 | Region detection and visual perception model training method, electronic equipment and storage medium |
| Publication number | Publication date |
|---|---|
| CN107169464B (en) | 2019-04-09 |
| Publication | Publication Date | Title |
|---|---|---|
| CN107169464A (en) | A kind of Method for Road Boundary Detection based on laser point cloud | |
| CN109858460B (en) | Lane line detection method based on three-dimensional laser radar | |
| CN112801022B (en) | Method for rapidly detecting and updating road boundary of unmanned mining card operation area | |
| CN111079611B (en) | Automatic extraction method for road surface and marking line thereof | |
| CN112200171B (en) | Road point cloud extraction method based on scanning lines | |
| CN104636763B (en) | A kind of road and obstacle detection method based on automatic driving car | |
| CN105160309B (en) | Three lanes detection method based on morphological image segmentation and region growing | |
| CN105740782B (en) | A quantification method of driver's lane changing process based on monocular vision | |
| CN104573627B (en) | Lane line based on bianry image retains and detection method | |
| CN111563412B (en) | Rapid lane line detection method based on parameter space voting and Bessel fitting | |
| CN113496491B (en) | Road surface segmentation method and device based on multi-line laser radar | |
| CN103295420A (en) | Method for recognizing lane line | |
| CN106503678A (en) | Roadmarking automatic detection and sorting technique based on mobile laser scanning point cloud | |
| KR102667741B1 (en) | Method and apparatus of displaying 3d object | |
| CN106650640A (en) | Negative obstacle detection method based on local structure feature of laser radar point cloud | |
| US8718329B2 (en) | Top-down view classification in clear path detection | |
| CN114488073A (en) | Method for processing point cloud data acquired by laser radar | |
| CN103020948A (en) | Night image characteristic extraction method in intelligent vehicle-mounted anti-collision pre-warning system | |
| CN104850834A (en) | Road boundary detection method based on three-dimensional laser radar | |
| KR101067437B1 (en) | Lane detection method and lane departure detection system using the same | |
| CN107895151A (en) | Method for detecting lane lines based on machine vision under a kind of high light conditions | |
| CN109190483B (en) | A Vision-Based Lane Line Detection Method | |
| CN114648654A (en) | A clustering method that fuses point cloud semantic categories and distances | |
| CN102201054A (en) | Method for detecting street lines based on robust statistics | |
| CN107256633A (en) | A kind of vehicle type classification method based on monocular cam three-dimensional estimation |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |